Tokenization and Chunking in Spring Boot

Implement recursive splitting that falls back through paragraph, sentence, word, and token boundaries while respecting a token budget.

Implement Markdown-aware splitting that preserves headings and fenced code blocks and records heading paths as chunk metadata.

Compare fixed-size, recursive, Markdown-aware, sentence-based, and semantic chunking for boundary quality, retrieval relevance, duplication, latency, and token cost.